Ensemble learning with explainable AI for improved heart disease prediction based on multiple datasets

Le résumé fourni par la source

Heart disease is a serious threat to human health; it is one of the leading causes of death. Being able to predict it in advance can help doctors separate patients into different categories based on risk levels and provide the most needed care to those who require it most. One of the ways to predict it is using the machine learning method. However, it should be noted that there are some issues with such models, such as overfitting, noise sensitivity, and sometimes a relatively low predictive performance. To address these limitations, this proposed study aims to develop an ensemble learning-based framework for improved heart disease prediction using multiple datasets. In this study, different machine learning classification algorithms will be implemented and evaluated. Ensemble techniques, particularly voting and stacking, will be investigated by combining the predictions of multiple diverse base classifiers The proposed study will adopt the following procedure: data pre-processing, selection of predictive features, data normalization, model training, hyperparameter optimization, and performance evaluation. The following performance metrics will be used to assess the models’ performance: accuracy, precision, recall, F1-score, specificity, and ROC-AUC. The methods mentioned above will be used to test, analyze, and confirm the reliability of results. Moreover, we will apply Explainable Artificial Intelligence (XAI) to evaluate model performance in terms of the clinical features that are most relevant in predicting heart disease. We will use SHAP (SHapley Additive exPlanations) as XAI for interpreting the results. The proposed framework is expected to improve prediction accuracy and robustness compared the proposed study examines the effectiveness of ensemble learning and explainable artificial intelligence algorithms to predict heart disease risks and provide decision-making support to healthcare practitioners.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.